A radar image jamming type identification and jamming degree evaluation method

Abstract

The application discloses a radar image interference type identification and interference degree evaluation method and belongs to the technical field of radar interference perception. The method comprises the following steps: S1, extracting the spatial domain statistical feature vector of the to-be-tested radar image through a spatial domain statistical feature vector extraction model; S2, extracting the frequency domain energy feature vector of the to-be-tested radar image through a frequency domain statistical feature vector extraction model; S3, identifying the interference type of the to-be-tested radar image, and completing interference type identification; and S4, performing pixel-level segmentation on the interference area of the to-be-tested radar image and confirming the interference degree, and completing interference degree evaluation.

Description

Technical Field

[0001] This invention belongs to the field of radar interference sensing technology, and in particular relates to a method for identifying radar image interference types and assessing interference levels. Background Technology

[0002] Currently, radar SAR imagery is increasingly widely used, gradually expanding its application scope in both military and civilian fields due to its all-weather, 24 / 7 advantages. However, it faces various interferences in complex environments, severely restricting the improvement of radar SAR image perception capabilities. There is an urgent need to further clarify the degree of interference in radar SAR images to eliminate adverse information and support an increase in the proportion of effective information in the radar SAR image perception process. Existing radar image interference identification and assessment methods mostly perform radio frequency interference detection at the echo level, lacking the ability to accurately identify multiple interference types at the image level. Furthermore, they are insufficient in assessing the degree of image interference, failing to quantitatively evaluate the extent of radar interference. Summary of the Invention

[0003] The purpose of this invention is to provide a method for identifying radar image interference types and assessing interference levels. It designs a process for interference level assessment driven by multi-dimensional feature extraction, multi-type interference identification, and pixel-level interference segmentation, which can accurately determine the usability of radar SAR images and effectively support the reasonable allocation of SAR image usage weights in application processes.

[0004] To achieve the above objectives, the present invention provides a method for identifying radar image interference types and assessing interference levels, comprising the following steps: S1. Extract the spatial statistical feature vector of the radar image under test through the spatial statistical feature vector extraction model; S2. Extract the frequency domain energy feature vector of the radar image under test through the frequency domain statistical feature vector extraction model; S3. Identify the type of interference in the radar image under test and complete the interference type identification. S4. Perform pixel-level segmentation and interference level confirmation of the interference area in the radar image under test, and complete the interference level assessment.

[0005] Preferably, the process of S1 is as follows: S11, Radar SAR image to be tested Perform L-level wavelet decomposition to obtain the sub-band coefficients of each level. ; in Indicates the first Horizontal detail sub-bands of the layer, Indicates the first Vertical detail sub-bands of the layer Indicates the first Diagonal detail sub-bands of the layer; S12, Calculate the first Mean value of sub-band coefficient at the layer level Standard deviation and kurtosis ; S13, Calculate the... Mean value of vertical sub-zone coefficient of layer Standard deviation With kurtosis ; S14, Calculate the... Mean of the coefficient of the diagonal sub-band Standard deviation With kurtosis ; S15. Construct a spatial statistical feature vector using the statistical characteristics of each sub-band in S12-S14. The process is as follows: .

[0006] Preferably, the calculation process for S12 is as follows: ; ; ; in Indicates the first Spatial support domain with layer band coefficient. This indicates the total number of pixels in the field.

[0007] Preferably, the calculation process for S13 is as follows: ; ; .

[0008] Preferably, the calculation process for S14 is as follows: ; ; .

[0009] Preferably, the process of S2 is as follows: S21. The radar SAR image to be tested Perform a two-dimensional fast Fourier transform to obtain the spectrum. The process is as follows: ; in, Indicates the image width. Indicates the image height; S22. Divide the spectrum into sector sections along the range and azimuth directions, and calculate the energy concentration of each section. The process is as follows: ; in Indicates the first A sector-based frequency domain partition; S23. Calculate the frequency domain energy eigenvector using energy concentration. .

[0010] Preferably, the process in S23 is as follows: S231. Calculate the main lobe energy ratio of the radar SAR image under test using the spectrum. The process is as follows: ; S232. Calculate the effective spectral bandwidth index of the radar SAR image to be tested using the spectrum. The process is as follows: ; S233, combining the energy concentration of each interval with the main lobe energy proportion and the effective spectral bandwidth index. The process is as follows: .

[0011] Preferably, the process of S3 is as follows: S31, obtained from splicing S1 and S2 respectively and To obtain splicing features The process is as follows: ; in Indicates feature concatenation operation; S32. Calculate intermediate features by splicing features. The process is as follows: ; in and All are learnable weight parameters. and All are learnable bias vectors. Represents a linear rectified function; S33. Calculate the interference identification loss function. The interference type identification process is as follows: ; in This indicates the Softmax operation. Let cross-entropy be the loss function. Interference type labels are manually assigned.

[0012] Preferably, the process of S4 is as follows: S41. Calculate the probability that the image in the radar image under test is predicted to be interference. The process is as follows: ; in For feature extractors, Represents a sigmoid function; S42. Use pixel-level Focal loss function as the segmentation loss function. The calculation process is as follows: ; in As a category balance factor, This is a focusing parameter used to reduce the weight of easily classified samples. For summation functions; S43. Calculate the interference mask of the radar image to be tested. The process is as follows: ; in For rounding operations, Indicates pixel position; S44. Calculate the degree of interference in the radar image under test. The process is as follows: ; in For the feature extractor in the interference level assessment model, This is a global pooling function. It is a multilayer perceptron; S45. Optimize the interference level model using the cross-entropy loss function, as follows: ; in The interference level labels are manually marked. The loss function represents the model for assessing the degree of interference. S46. Based on preset thresholds, the interference level is divided into three levels: mild, moderate, and severe, to complete the interference level assessment of radar SAR images.

[0013] Preferably, the radar image interference type identification and interference level assessment method is implemented through a radar image interference type identification and interference level assessment system, which includes a multi-scale spatial and frequency domain feature extraction module, an interference type identification module, and an interference semantic segmentation and interference level assessment module. The multi-scale spatial and frequency domain feature extraction module includes a spatial domain statistical feature vector extraction model and a frequency domain statistical feature vector extraction model. It extracts the spatial and frequency domain features of radar SAR images through wavelet transform and fast Fourier transform to support subsequent interference type identification. The interference type identification module includes an interference type identification model, which combines the spatial and frequency domain features of radar SAR images to identify the interference type in the images. The interference semantic segmentation and interference level assessment module includes an interference semantic segmentation model and an interference level assessment model. It uses a pixel-level segmentation algorithm to identify interference regions in the image at the pixel level, and uses this as a basis to assess the degree of interference in conjunction with the original radar SAR image.

[0014] Therefore, the present invention adopts the above-mentioned method for identifying radar image interference type and assessing interference level. By designing a hierarchical model for multi-scale spatial and frequency domain feature extraction, interference type identification, interference semantic segmentation and interference level assessment, it can accurately identify the interference type in radar SAR images in complex scenarios, and ensure accurate perception of the degree of interference in the image by extracting pixel-level interference regions.

[0015] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0016] Figure 1 This is an overall flowchart of a radar image interference type identification and interference level assessment method according to the present invention; Figure 2 This is a block diagram of a radar image interference type identification and interference level assessment method according to the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Specific model specifications need to be selected and determined according to the actual specifications of the device, etc. The specific selection calculation method adopts existing technology in the art, and therefore will not be described in detail.

[0018] Example like Figure 1 and Figure 2 As shown, the present invention provides a method for identifying radar image interference types and assessing interference levels, comprising the following steps: The multi-scale spatial-frequency domain feature extraction module performs multi-scale spatial-frequency domain joint feature extraction on the input radar SAR image to obtain a spatial statistical feature vector. With frequency domain energy eigenvectors ; S1. Extract the spatial statistical feature vector of the radar image under test through the spatial statistical feature vector extraction model; S11. Spatial Statistical Feature Vector Extraction Model for Radar SAR Images Under Test Perform L-level wavelet decomposition to obtain the sub-band coefficients of each level. ; in Indicates the first Horizontal detail sub-bands of the layer, Indicates the first Vertical detail sub-bands of the layer Indicates the first Diagonal detail sub-bands of the layer; S12, Calculate the first Mean value of sub-band coefficient at the layer level Standard deviation and kurtosis ; ; ; ; in Indicates the first Spatial support domain with layer band coefficient. This indicates the total number of pixels in the field.

[0019] S13, Calculate the... Mean value of vertical sub-zone coefficient of layer Standard deviation With kurtosis ; ; ; .

[0020] S14, Calculate the... Mean of the coefficient of the diagonal sub-band Standard deviation With kurtosis ; ; ; .

[0021] S15. Construct a spatial statistical feature vector using the statistical characteristics of each sub-band in S12-S14. The process is as follows: .

[0022] S2. Extract the frequency domain energy feature vector of the radar image under test through the frequency domain statistical feature vector extraction model; S21. Frequency domain statistical feature vector extraction model for the radar SAR image under test. Perform a two-dimensional fast Fourier transform to obtain the spectrum. The process is as follows: ; in, Indicates the image width. Indicates the image height; S22. Divide the spectrum into sector sections along the range and azimuth directions, and calculate the energy concentration of each section. The process is as follows: ; in Indicates the first A sector-based frequency domain partition; S23. Calculate the frequency domain energy eigenvector using energy concentration. .

[0023] S231. Calculate the main lobe energy ratio of the radar SAR image under test using the spectrum. The process is as follows: ; S232. Calculate the effective spectral bandwidth index of the radar SAR image to be tested using the spectrum. The process is as follows: ; S233, combining the energy concentration of each interval with the main lobe energy proportion and the effective spectral bandwidth index. The process is as follows: .

[0024] S3. Identify the type of interference in the radar image under test and complete the interference type identification. The interference type identification module utilizes joint features of the spatial and frequency domains to design an interference type identification model and output interference type identification results. It identifies radio frequency interference, speckle noise, range / azimuth ambiguity, sidelobe interference, and spoofing interference. The model is trained and optimized using the cross-entropy loss function.

[0025] S31, obtained from splicing S1 and S2 respectively and To obtain splicing features The process is as follows: ; in Indicates feature concatenation operation; S32. Calculate intermediate features by splicing features. The process is as follows: ; in and All are learnable weight parameters. and All are learnable bias vectors. Represents a linear rectified function; S33. Calculate the interference identification loss function. The interference type identification process is as follows: ; in This indicates the Softmax operation. Let cross-entropy be the loss function. Interference type labels are manually assigned.

[0026] S4. Perform pixel-level segmentation and interference level confirmation of the interference area in the radar image under test, and complete the interference level assessment.

[0027] The interference semantic segmentation and interference degree assessment module extracts pixel-level interference regions through a semantic segmentation network to form an interference region mask, and maps the degree of interference through mask statistics. S41. Calculate the probability that the image in the radar image under test is predicted to be interference. The process is as follows: ; in For feature extractors, Represents a sigmoid function; S42. The interference semantic segmentation model uses the pixel-level Focal loss function as the segmentation loss function. The calculation process is as follows: ; in As a category balance factor, This is a focusing parameter used to reduce the weight of easily classified samples. For summation functions; The interference level assessment model uses pixel-level interference masks and radar SAR images to assess the degree of interference, and classifies the interference level into three levels: mild, moderate and severe according to preset thresholds. The model is optimized using the cross-entropy loss function. S43. Calculate the interference mask of the radar image to be tested. The process is as follows: ; in For rounding operations, Indicates pixel position; S44. Calculate the degree of interference in the radar image under test. The process is as follows: ; in For the feature extractor in the interference level assessment model, This is a global pooling function. It is a multilayer perceptron; S45. Optimize the interference level model using the cross-entropy loss function, as follows: ; in The interference level labels are manually marked. The loss function represents the model for assessing the degree of interference. S46. Based on preset thresholds, the interference level is divided into three levels: mild, moderate, and severe, to complete the interference level assessment of radar SAR images.

[0028] The radar image interference type identification and interference level assessment method is implemented through a radar image interference type identification and interference level assessment system, including a multi-scale spatial and frequency domain feature extraction module, an interference type identification module, and an interference semantic segmentation and interference level assessment module. The multi-scale spatial and frequency domain feature extraction module includes a spatial domain statistical feature vector extraction model and a frequency domain statistical feature vector extraction model. It extracts the spatial and frequency domain features of radar SAR images through wavelet transform and fast Fourier transform to support subsequent interference type identification. The interference type identification module includes an interference type identification model, which combines the spatial and frequency domain features of radar SAR images to identify the interference type in the images. The interference semantic segmentation and interference level assessment module includes an interference semantic segmentation model and an interference level assessment model. It uses a pixel-level segmentation algorithm to identify interference regions in the image at the pixel level, and uses this as a basis to assess the degree of interference in conjunction with the original radar SAR image.

[0029] Therefore, this invention employs a radar image interference type identification and interference level assessment method. Based on radar SAR images, it performs multi-scale spatial and frequency domain feature extraction, interference type identification, and interference semantic segmentation. Finally, it assesses the interference level to generate an image interference complexity identification result. First, the input radar SAR image is decomposed into multi-resolution features, extracting spatial statistical features and frequency domain energy distribution features respectively. Then, based on a feature fusion network, multiple types of interference are identified. On this basis, pixel-level interference localization is performed to form an interference mask. Finally, an interference level assessment model is used to calculate the degree of interference in the image, supporting the improvement of radar SAR image perception capabilities in complex scenarios.

[0030] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for identifying radar image interference types and assessing interference levels, characterized in that, Includes the following steps: S1. Extract the spatial statistical feature vector of the radar image under test using the spatial statistical feature vector extraction model. The process is as follows: S11, Radar SAR image to be tested Perform L-level wavelet decomposition to obtain the sub-band coefficients of each level. ; in Indicates the first Horizontal detail sub-bands of the layer Indicates the first Vertical detail sub-bands of the layer Indicates the first Diagonal detail sub-bands of the layer; S12, Calculate the first Mean value of sub-band coefficient at the layer level Standard deviation and kurtosis ; S13, Calculate the... Mean value of vertical sub-zone coefficient of layer Standard deviation With kurtosis ; S14, Calculate the... Mean of the coefficient of the diagonal sub-band Standard deviation With kurtosis ; S15. Construct a spatial statistical feature vector using the statistical characteristics of each sub-band in S12-S14. The process is as follows: ； S2. Extract the frequency domain energy feature vector of the radar image under test using the frequency domain statistical feature vector extraction model. The process is as follows: S21. The radar SAR image to be tested Perform a two-dimensional fast Fourier transform to obtain the spectrum. The process is as follows: ； in, Indicates the image width. Indicates the image height; S22. Divide the spectrum into sector sections along the range and azimuth directions, and calculate the energy concentration of each section. The process is as follows: ； in Indicates the first A sector-based frequency domain partition; S23. Calculate the frequency domain energy eigenvector using energy concentration. ; S3, obtained from splicing S1 and S2 respectively and To obtain splicing features Based on the stitching feature, the interference type of the radar image under test is identified, and the interference type identification is completed. S4. Perform pixel-level segmentation and interference level confirmation of the interference area in the radar image under test, and complete the interference level assessment.

2. The method for identifying radar image interference types and assessing interference levels according to claim 1, characterized in that, The calculation process for S12 is as follows: ； ； ； in Indicates the first Spatial support domain with layer band coefficient. This indicates the total number of pixels in the field.

3. The method for identifying radar image interference types and assessing interference levels according to claim 2, characterized in that, The calculation process for S13 is as follows: ； ； 。 4. The method for identifying radar image interference types and assessing interference levels according to claim 3, characterized in that, The calculation process for S14 is as follows: ； ； 。 5. The radar image interference type identification and interference degree assessment method according to claim 4, characterized in that, The process of S23 is as follows: S231. Calculate the main lobe energy ratio of the radar SAR image under test using the spectrum. The process is as follows: ； S232. Calculate the effective spectral bandwidth index of the radar SAR image to be tested using the spectrum. The process is as follows: ； S233, combining the energy concentration of each interval with the main lobe energy proportion and the effective spectral bandwidth index. The process is as follows: 。 6. The radar image interference type identification and interference degree assessment method according to claim 5, characterized in that, The process of S3 is as follows: S31, obtained from splicing S1 and S2 respectively and To obtain splicing features The process is as follows: ； in Indicates feature concatenation operation; S32. Calculate intermediate features by splicing features. The process is as follows: ； in and All are learnable weight parameters. and All are learnable bias vectors. Represents a linear rectified function; S33. Calculate the interference identification loss function. The interference type identification process is as follows: ； in This indicates the Softmax operation. Let cross-entropy be the loss function. Interference type labels are manually assigned.

7. The radar image interference type identification and interference degree assessment method according to claim 6, characterized in that, The process of S4 is as follows: S41. Calculate the probability that the image in the radar image under test is predicted to be interference. The process is as follows: ； in For feature extractors, Represents a sigmoid function; S42. Use pixel-level Focal loss function as the segmentation loss function. The calculation process is as follows: ； in As a category balance factor, This is a focusing parameter used to reduce the weight of easily classified samples. For summation functions; S43. Calculate the interference mask of the radar image under test. The process is as follows: ； in For rounding operations, Indicates pixel position; S44. Calculate the degree of interference in the radar image under test. The process is as follows: ； in For the feature extractor in the interference level assessment model, This is a global pooling function. It is a multilayer perceptron; S45. Optimize the interference level model using the cross-entropy loss function, as follows: ； in The interference level labels are manually marked. The loss function represents the model for assessing the degree of interference. S46. Based on preset thresholds, the interference level is divided into three levels: mild, moderate, and severe, to complete the interference level assessment of radar SAR images.

8. The method for identifying radar image interference types and assessing interference levels according to claim 7, characterized in that, The radar image interference type identification and interference level assessment method is implemented through a radar image interference type identification and interference level assessment system, including a multi-scale spatial and frequency domain feature extraction module, an interference type identification module, and an interference semantic segmentation and interference level assessment module. The multi-scale spatial and frequency domain feature extraction module includes a spatial domain statistical feature vector extraction model and a frequency domain statistical feature vector extraction model. It extracts the spatial and frequency domain features of radar SAR images through wavelet transform and fast Fourier transform to support subsequent interference type identification. The interference type identification module includes an interference type identification model, which combines the spatial and frequency domain features of radar SAR images to identify the interference type in the images. The interference semantic segmentation and interference level assessment module includes an interference semantic segmentation model and an interference level assessment model. It uses a pixel-level segmentation algorithm to identify interference regions in the image at the pixel level, and uses this as a basis to assess the degree of interference in conjunction with the original radar SAR image.

Images